Proteomics Profiling of Osteoporosis and Osteopenia Patients and Associated Network Analysis

Abstract

Bone mass reduction due to an imbalance in osteogenesis and osteolysis is characterized by low bone mineral density (LBMD) and is clinically classified as osteopenia (ON) or osteoporosis (OP), which is more severe. Multiple biomarkers for diagnosing OP and its progression have been reported; however, most of these lack specificity. This cohort study aimed to investigate sensitive and specific LBMD-associated protein biomarkers in patients diagnosed with ON and OP. A label-free liquid chromatography-mass spectrometry (LC-MS) proteomics approach was used to analyze serum samples. Patients' proteomics profiles were filtered for potential confounding effects, such as age, sex, chronic diseases, and medication. A distinctive proteomics profile between the control, ON, and OP groups (Q² = 0.7295, R² = 0.9180) was identified, and significant dysregulation in a panel of proteins (n = 20) was common among the three groups. A comparison of these proteins showed that the levels of eight proteins were upregulated in ON, compared to those in the control and the OP groups, while the levels of eleven proteins were downregulated in the ON group compared to those in the control group. Interestingly, only one protein, myosin heavy chain 14 (MYH14), showed a linear increase from the control to the ON group, with the highest abundance in the OP group. A significant separation in the proteomics profile between the ON and OP groups (Q² = 0.8760, R² = 0.991) was also noted. Furthermore, a total of twenty-six proteins were found to be dysregulated between the ON and the OP groups, with fourteen upregulated and twelve downregulated proteins in the OP, compared to that in the ON group. Most of the identified dysregulated proteins were immunoglobulins, complement proteins, cytoskeletal proteins, coagulation factors, and various enzymes. Of these identified proteins, the highest area under the curve (AUC) in the receiver operating characteristic (ROC) analysis was related to three proteins (immunoglobulin Lambda constant 1 (IGLC1), RNA binding protein (MEX3B), and fibulin 1 (FBLN1)). Multiple reaction monitoring (MRM), LC-MS, was used to validate some of the identified proteins. A network pathway analysis of the differentially abundant proteins demonstrated dysregulation of inflammatory signaling pathways in the LBMD patients, including the tumor necrosis factor (TNF), toll-like receptor (TL4), and interferon-γ (IFNG) signaling pathways. These results reveal the existence of potentially sensitive protein biomarkers that could be used in further investigations of bone health and OP progression.

Keywords: bone mineral density (BMD); mass spectrometry; osteopenia; osteoporosis; proteomics.

Figure 1

Overall proteomics analysis and exclusion of confounders-associated proteins. (A) Determination of confounder-independent proteins from the overall detected proteins. (B) Venn diagram demonstrating overlap between confounder-independent proteins (medications, TD, Sex, and T2DM) (n = 212, 154, 122, and 123, respectively) using moderate t-test and considering fold change (FC 1.5) and cut-off p-value < 0.05. A total of 68 proteins were identified as being significantly associated with LBMD, independent of the effect of confounders. Abbreviations: LBMD: low bone mineral density; TD: thyroid disease; T2DM: type 2 diabetes mellitus. * Two-way ANOVA with FDR-corrected p-value (FDRp) cutoff = 0.05.

Figure 2

Proteomics profiling between healthy control (Ctrl), osteopenic (ON), and osteoporotic (OP) patients. (A) Orthogonal partial least squares-discriminant analysis (OPLS-DA) cross-validation illustrates the significant differences between the three study groups (Ctrl, ON, OP) (Q² = 0.7295, and R² = 0.9180). (B) Venn diagram demonstrating the significantly dysregulated proteins between Ctrl vs. ON (n = 65), Ctrl vs. OP (n = 34), and ON vs. OP (n = 51), considering FC of 1.5 and p-value of 0.05. Also shown is the identification of common and significant proteins (n = 20) between the three study groups. (C) Levels of commonly dysregulated proteins (G20) among the three groups, where G8 were upregulated to the highest abundance in the ON group, compared to that in the control group and then downregulated in the OP group, compared to that in the ON group, while G11 was downregulated to the lowest abundance in ON, compared to that in control and then upregulated in OP compared to that in ON. (D) Heat map showing the identity and expression levels of the 20 significantly detected proteins among the three study groups and also those associated with fracture history (FH) (highlighted with an asterisk). Green and colors mean down and up-regulation, respectively.

Figure 2

Proteomics profiling between healthy control (Ctrl), osteopenic (ON), and osteoporotic (OP) patients. (A) Orthogonal partial least squares-discriminant analysis (OPLS-DA) cross-validation illustrates the significant differences between the three study groups (Ctrl, ON, OP) (Q² = 0.7295, and R² = 0.9180). (B) Venn diagram demonstrating the significantly dysregulated proteins between Ctrl vs. ON (n = 65), Ctrl vs. OP (n = 34), and ON vs. OP (n = 51), considering FC of 1.5 and p-value of 0.05. Also shown is the identification of common and significant proteins (n = 20) between the three study groups. (C) Levels of commonly dysregulated proteins (G20) among the three groups, where G8 were upregulated to the highest abundance in the ON group, compared to that in the control group and then downregulated in the OP group, compared to that in the ON group, while G11 was downregulated to the lowest abundance in ON, compared to that in control and then upregulated in OP compared to that in ON. (D) Heat map showing the identity and expression levels of the 20 significantly detected proteins among the three study groups and also those associated with fracture history (FH) (highlighted with an asterisk). Green and colors mean down and up-regulation, respectively.

Figure 3

Proteomics profiling between ON and OP patients. (A) Orthogonal partial least squares-discriminant analysis (OPLS-DA) score plot showing the relative separation between ON and OP groups (Q² = 0.876, and R² = 0.991) after excluding three outlier values detected using the random forest algorithm. (B) Volcano plot analysis of ON versus OP showing significantly dysregulated proteins (false discovery rate (FDR)-corrected p-value < 0.05, and fold change (FC) > 1.5 or < 0.67). A total of (G49) proteins were found to be dysregulated (26 up-regulated and 23 down-regulated) in OP patients, compared to those in ON patients. (C) Venn diagram illustrating an overlap between the confounder’s independent proteins (n = 68) and the dysregulated proteins between the ON and OP groups (G49). A total of 26 proteins were significantly dysregulated (14 up-regulated and 12 down-regulated) in OP, compared to those in ON patients. (D) Heat map showing the expression and the identity of the dysregulated proteins between the ON and OP groups along with fracture history (FH)-associated proteins (highlighted with an asterisk). Green and colors mean down and up-regulation, respectively.

Figure 3

Proteomics profiling between ON and OP patients. (A) Orthogonal partial least squares-discriminant analysis (OPLS-DA) score plot showing the relative separation between ON and OP groups (Q² = 0.876, and R² = 0.991) after excluding three outlier values detected using the random forest algorithm. (B) Volcano plot analysis of ON versus OP showing significantly dysregulated proteins (false discovery rate (FDR)-corrected p-value < 0.05, and fold change (FC) > 1.5 or < 0.67). A total of (G49) proteins were found to be dysregulated (26 up-regulated and 23 down-regulated) in OP patients, compared to those in ON patients. (C) Venn diagram illustrating an overlap between the confounder’s independent proteins (n = 68) and the dysregulated proteins between the ON and OP groups (G49). A total of 26 proteins were significantly dysregulated (14 up-regulated and 12 down-regulated) in OP, compared to those in ON patients. (D) Heat map showing the expression and the identity of the dysregulated proteins between the ON and OP groups along with fracture history (FH)-associated proteins (highlighted with an asterisk). Green and colors mean down and up-regulation, respectively.

Figure 4

Results of biomarker evaluation in ON and OP. (A) Exploratory ROC curve generated by the OPLS-DA model; AUC values were calculated by mathematical integration of the combination of 5, 10, 15, 25, 50, and 100 proteins. (B–D) Three proteins with the highest AUC: (B) IGLC1, AUC = 0.929; (C) MEX3B, AUC = 0.884; and (D) FBLNI, AUC = 0.883.

Figure 5

Validation of the expression of selected proteins in the ON and OP groups using multiple reaction monitoring (MRM). (A) Representative MRM chromatograms for protein signature peptides selected from the Skyline Software and confirmed using PeptideAtlas. (B) Scatter plots with bar-graph for the expressions of selected proteins. The differences between the study groups were evaluated using an unpaired t-test with significance set at p-value <0.01 (denoted by **), <0.001 (denoted by ***), and <0.0001(denoted by ****).

Figure 6

Network analysis and biological pathways related to the significantly identified proteins in the study population. (A) Network pathway analysis of the significantly dysregulated proteins identified in the ON group, compared to those in the OP group revealed that they were related to the developmental disorder, hereditary disorder, and metabolic disease. The analysis also showed the involvement of the TNF and IFNG signaling pathways. (B) The top canonical pathways related to the significantly dysregulated proteins identified in the ON group, compared to those in the OP group.